Linux-libre 4.4-gnupck-4.4-gnu

34 files changed, 451 insertions, 218 deletions

diff --git a/arch/sparc/include/asm/atomic_64.h b/arch/sparc/include/asm/atomic_64.h
index 917084ace..f2fbf9e16 100644
--- a/arch/sparc/include/asm/atomic_64.h
+++ b/arch/sparc/include/asm/atomic_64.h
@@ -14,11 +14,11 @@
 #define ATOMIC_INIT(i)		{ (i) }
 #define ATOMIC64_INIT(i)	{ (i) }
 
-#define atomic_read(v)		ACCESS_ONCE((v)->counter)
-#define atomic64_read(v)	ACCESS_ONCE((v)->counter)
+#define atomic_read(v)		READ_ONCE((v)->counter)
+#define atomic64_read(v)	READ_ONCE((v)->counter)
 
-#define atomic_set(v, i)	(((v)->counter) = i)
-#define atomic64_set(v, i)	(((v)->counter) = i)
+#define atomic_set(v, i)	WRITE_ONCE(((v)->counter), (i))
+#define atomic64_set(v, i)	WRITE_ONCE(((v)->counter), (i))
 
 #define ATOMIC_OP(op)							\
 void atomic_##op(int, atomic_t *);					\
diff --git a/arch/sparc/include/asm/elf_64.h b/arch/sparc/include/asm/elf_64.h
index 370ca1e71..93310837c 100644
--- a/arch/sparc/include/asm/elf_64.h
+++ b/arch/sparc/include/asm/elf_64.h
@@ -95,6 +95,7 @@
  * really available.  So we simply advertise only "crypto" support.
  */
 #define HWCAP_SPARC_CRYPTO	0x04000000 /* CRYPTO insns available */
+#define HWCAP_SPARC_ADI		0x08000000 /* ADI available */
 
 #define CORE_DUMP_USE_REGSET
 
diff --git a/arch/sparc/include/asm/topology_64.h b/arch/sparc/include/asm/topology_64.h
index 01d170462..bec481aac 100644
--- a/arch/sparc/include/asm/topology_64.h
+++ b/arch/sparc/include/asm/topology_64.h
@@ -31,6 +31,9 @@ static inline int pcibus_to_node(struct pci_bus *pbus)
 	 cpu_all_mask : \
 	 cpumask_of_node(pcibus_to_node(bus)))
 
+int __node_distance(int, int);
+#define node_distance(a, b) __node_distance(a, b)
+
 #else /* CONFIG_NUMA */
 
 #include <asm-generic/topology.h>
diff --git a/arch/sparc/include/uapi/asm/asi.h b/arch/sparc/include/uapi/asm/asi.h
index aace6f313..7ad7203de 100644
--- a/arch/sparc/include/uapi/asm/asi.h
+++ b/arch/sparc/include/uapi/asm/asi.h
@@ -279,7 +279,7 @@
 				      * Most-Recently-Used, primary,
 				      * implicit
 				      */
-#define ASI_ST_BLKINIT_MRU_S	0xf2 /* (NG4) init-store, twin load,
+#define ASI_ST_BLKINIT_MRU_S	0xf3 /* (NG4) init-store, twin load,
 				      * Most-Recently-Used, secondary,
 				      * implicit
 				      */
diff --git a/arch/sparc/include/uapi/asm/mman.h b/arch/sparc/include/uapi/asm/mman.h
index 0b14df33c..9765896ec 100644
--- a/arch/sparc/include/uapi/asm/mman.h
+++ b/arch/sparc/include/uapi/asm/mman.h
@@ -17,6 +17,7 @@
 
 #define MCL_CURRENT     0x2000          /* lock all currently mapped pages */
 #define MCL_FUTURE      0x4000          /* lock all additions to address space */
+#define MCL_ONFAULT	0x8000		/* lock all pages that are faulted in */
 
 #define MAP_POPULATE	0x8000		/* populate (prefault) pagetables */
 #define MAP_NONBLOCK	0x10000		/* do not block on IO */
diff --git a/arch/sparc/include/uapi/asm/unistd.h b/arch/sparc/include/uapi/asm/unistd.h
index 6f35f4df1..1c26d440d 100644
--- a/arch/sparc/include/uapi/asm/unistd.h
+++ b/arch/sparc/include/uapi/asm/unistd.h
@@ -416,8 +416,14 @@
 #define __NR_memfd_create	348
 #define __NR_bpf		349
 #define __NR_execveat		350
+#define __NR_membarrier		351
+#define __NR_userfaultfd	352
+#define __NR_bind		353
+#define __NR_listen		354
+#define __NR_setsockopt		355
+#define __NR_mlock2		356
 
-#define NR_syscalls		351
+#define NR_syscalls		357
 
 /* Bitmask values returned from kern_features system call.  */
 #define KERN_FEATURE_MIXED_MODE_STACK	0x00000001
diff --git a/arch/sparc/kernel/head_64.S b/arch/sparc/kernel/head_64.S
index 3d61fcae7..f2d30cab5 100644
--- a/arch/sparc/kernel/head_64.S
+++ b/arch/sparc/kernel/head_64.S
@@ -946,6 +946,12 @@ ENTRY(__retl_one)
 	 mov	1, %o0
 ENDPROC(__retl_one)
 
+ENTRY(__retl_one_fp)
+	VISExitHalf
+	retl
+	 mov	1, %o0
+ENDPROC(__retl_one_fp)
+
 ENTRY(__ret_one_asi)
 	wr	%g0, ASI_AIUS, %asi
 	ret
@@ -958,6 +964,13 @@ ENTRY(__retl_one_asi)
 	 mov	1, %o0
 ENDPROC(__retl_one_asi)
 
+ENTRY(__retl_one_asi_fp)
+	wr	%g0, ASI_AIUS, %asi
+	VISExitHalf
+	retl
+	 mov	1, %o0
+ENDPROC(__retl_one_asi_fp)
+
 ENTRY(__retl_o1)
 	retl
 	 mov	%o1, %o0
diff --git a/arch/sparc/kernel/iommu.c b/arch/sparc/kernel/iommu.c
index 5320689c0..37686828c 100644
--- a/arch/sparc/kernel/iommu.c
+++ b/arch/sparc/kernel/iommu.c
@@ -161,7 +161,7 @@ static inline iopte_t *alloc_npages(struct device *dev,
 
 	entry = iommu_tbl_range_alloc(dev, &iommu->tbl, npages, NULL,
 				      (unsigned long)(-1), 0);
-	if (unlikely(entry == DMA_ERROR_CODE))
+	if (unlikely(entry == IOMMU_ERROR_CODE))
 		return NULL;
 
 	return iommu->page_table + entry;
@@ -253,7 +253,7 @@ static void dma_4u_free_coherent(struct device *dev, size_t size,
 	npages = IO_PAGE_ALIGN(size) >> IO_PAGE_SHIFT;
 	iommu = dev->archdata.iommu;
 
-	iommu_tbl_range_free(&iommu->tbl, dvma, npages, DMA_ERROR_CODE);
+	iommu_tbl_range_free(&iommu->tbl, dvma, npages, IOMMU_ERROR_CODE);
 
 	order = get_order(size);
 	if (order < 10)
@@ -426,7 +426,7 @@ static void dma_4u_unmap_page(struct device *dev, dma_addr_t bus_addr,
 	iommu_free_ctx(iommu, ctx);
 	spin_unlock_irqrestore(&iommu->lock, flags);
 
-	iommu_tbl_range_free(&iommu->tbl, bus_addr, npages, DMA_ERROR_CODE);
+	iommu_tbl_range_free(&iommu->tbl, bus_addr, npages, IOMMU_ERROR_CODE);
 }
 
 static int dma_4u_map_sg(struct device *dev, struct scatterlist *sglist,
@@ -492,7 +492,7 @@ static int dma_4u_map_sg(struct device *dev, struct scatterlist *sglist,
 					      &handle, (unsigned long)(-1), 0);
 
 		/* Handle failure */
-		if (unlikely(entry == DMA_ERROR_CODE)) {
+		if (unlikely(entry == IOMMU_ERROR_CODE)) {
 			if (printk_ratelimit())
 				printk(KERN_INFO "iommu_alloc failed, iommu %p paddr %lx"
 				       " npages %lx\n", iommu, paddr, npages);
@@ -571,7 +571,7 @@ iommu_map_failed:
 				iopte_make_dummy(iommu, base + j);
 
 			iommu_tbl_range_free(&iommu->tbl, vaddr, npages,
-					     DMA_ERROR_CODE);
+					     IOMMU_ERROR_CODE);
 
 			s->dma_address = DMA_ERROR_CODE;
 			s->dma_length = 0;
@@ -648,7 +648,7 @@ static void dma_4u_unmap_sg(struct device *dev, struct scatterlist *sglist,
 			iopte_make_dummy(iommu, base + i);
 
 		iommu_tbl_range_free(&iommu->tbl, dma_handle, npages,
-				     DMA_ERROR_CODE);
+				     IOMMU_ERROR_CODE);
 		sg = sg_next(sg);
 	}
 
diff --git a/arch/sparc/kernel/ldc.c b/arch/sparc/kernel/ldc.c
index 1ae5eb1bb..59d503866 100644
--- a/arch/sparc/kernel/ldc.c
+++ b/arch/sparc/kernel/ldc.c
@@ -1953,7 +1953,7 @@ static struct ldc_mtable_entry *alloc_npages(struct ldc_iommu *iommu,
 
 	entry = iommu_tbl_range_alloc(NULL, &iommu->iommu_map_table,
 				      npages, NULL, (unsigned long)-1, 0);
-	if (unlikely(entry < 0))
+	if (unlikely(entry == IOMMU_ERROR_CODE))
 		return NULL;
 
 	return iommu->page_table + entry;
diff --git a/arch/sparc/kernel/pci.c b/arch/sparc/kernel/pci.c
index b91d7f146..badf0951d 100644
--- a/arch/sparc/kernel/pci.c
+++ b/arch/sparc/kernel/pci.c
@@ -185,8 +185,10 @@ static unsigned long pci_parse_of_flags(u32 addr0)
 
 	if (addr0 & 0x02000000) {
 		flags = IORESOURCE_MEM | PCI_BASE_ADDRESS_SPACE_MEMORY;
-		flags |= (addr0 >> 22) & PCI_BASE_ADDRESS_MEM_TYPE_64;
 		flags |= (addr0 >> 28) & PCI_BASE_ADDRESS_MEM_TYPE_1M;
+		if (addr0 & 0x01000000)
+			flags |= IORESOURCE_MEM_64
+				 | PCI_BASE_ADDRESS_MEM_TYPE_64;
 		if (addr0 & 0x40000000)
 			flags |= IORESOURCE_PREFETCH
 				 | PCI_BASE_ADDRESS_MEM_PREFETCH;
@@ -655,6 +657,9 @@ struct pci_bus *pci_scan_one_pbm(struct pci_pbm_info *pbm,
 				pbm->io_space.start);
 	pci_add_resource_offset(&resources, &pbm->mem_space,
 				pbm->mem_space.start);
+	if (pbm->mem64_space.flags)
+		pci_add_resource_offset(&resources, &pbm->mem64_space,
+					pbm->mem_space.start);
 	pbm->busn.start = pbm->pci_first_busno;
 	pbm->busn.end	= pbm->pci_last_busno;
 	pbm->busn.flags	= IORESOURCE_BUS;
diff --git a/arch/sparc/kernel/pci_common.c b/arch/sparc/kernel/pci_common.c
index 944a06536..33524c1d5 100644
--- a/arch/sparc/kernel/pci_common.c
+++ b/arch/sparc/kernel/pci_common.c
@@ -406,6 +406,7 @@ void pci_determine_mem_io_space(struct pci_pbm_info *pbm)
 	}
 
 	num_pbm_ranges = i / sizeof(*pbm_ranges);
+	memset(&pbm->mem64_space, 0, sizeof(struct resource));
 
 	for (i = 0; i < num_pbm_ranges; i++) {
 		const struct linux_prom_pci_ranges *pr = &pbm_ranges[i];
@@ -451,7 +452,12 @@ void pci_determine_mem_io_space(struct pci_pbm_info *pbm)
 			break;
 
 		case 3:
-			/* XXX 64-bit MEM handling XXX */
+			/* 64-bit MEM handling */
+			pbm->mem64_space.start = a;
+			pbm->mem64_space.end = a + size - 1UL;
+			pbm->mem64_space.flags = IORESOURCE_MEM;
+			saw_mem = 1;
+			break;
 
 		default:
 			break;
@@ -465,15 +471,22 @@ void pci_determine_mem_io_space(struct pci_pbm_info *pbm)
 		prom_halt();
 	}
 
-	printk("%s: PCI IO[%llx] MEM[%llx]\n",
+	printk("%s: PCI IO[%llx] MEM[%llx]",
 	       pbm->name,
 	       pbm->io_space.start,
 	       pbm->mem_space.start);
+	if (pbm->mem64_space.flags)
+		printk(" MEM64[%llx]",
+		       pbm->mem64_space.start);
+	printk("\n");
 
 	pbm->io_space.name = pbm->mem_space.name = pbm->name;
+	pbm->mem64_space.name = pbm->name;
 
 	request_resource(&ioport_resource, &pbm->io_space);
 	request_resource(&iomem_resource, &pbm->mem_space);
+	if (pbm->mem64_space.flags)
+		request_resource(&iomem_resource, &pbm->mem64_space);
 
 	pci_register_legacy_regions(&pbm->io_space,
 				    &pbm->mem_space);
diff --git a/arch/sparc/kernel/pci_impl.h b/arch/sparc/kernel/pci_impl.h
index 75803c780..37222ca84 100644
--- a/arch/sparc/kernel/pci_impl.h
+++ b/arch/sparc/kernel/pci_impl.h
@@ -97,6 +97,7 @@ struct pci_pbm_info {
 	/* PBM I/O and Memory space resources. */
 	struct resource			io_space;
 	struct resource			mem_space;
+	struct resource			mem64_space;
 	struct resource			busn;
 
 	/* Base of PCI Config space, can be per-PBM or shared. */
diff --git a/arch/sparc/kernel/pci_sun4v.c b/arch/sparc/kernel/pci_sun4v.c
index d2fe57dad..836e8cef4 100644
--- a/arch/sparc/kernel/pci_sun4v.c
+++ b/arch/sparc/kernel/pci_sun4v.c
@@ -159,7 +159,7 @@ static void *dma_4v_alloc_coherent(struct device *dev, size_t size,
 	entry = iommu_tbl_range_alloc(dev, &iommu->tbl, npages, NULL,
 				      (unsigned long)(-1), 0);
 
-	if (unlikely(entry == DMA_ERROR_CODE))
+	if (unlikely(entry == IOMMU_ERROR_CODE))
 		goto range_alloc_fail;
 
 	*dma_addrp = (iommu->tbl.table_map_base + (entry << IO_PAGE_SHIFT));
@@ -187,7 +187,7 @@ static void *dma_4v_alloc_coherent(struct device *dev, size_t size,
 	return ret;
 
 iommu_map_fail:
-	iommu_tbl_range_free(&iommu->tbl, *dma_addrp, npages, DMA_ERROR_CODE);
+	iommu_tbl_range_free(&iommu->tbl, *dma_addrp, npages, IOMMU_ERROR_CODE);
 
 range_alloc_fail:
 	free_pages(first_page, order);
@@ -226,7 +226,7 @@ static void dma_4v_free_coherent(struct device *dev, size_t size, void *cpu,
 	devhandle = pbm->devhandle;
 	entry = ((dvma - iommu->tbl.table_map_base) >> IO_PAGE_SHIFT);
 	dma_4v_iommu_demap(&devhandle, entry, npages);
-	iommu_tbl_range_free(&iommu->tbl, dvma, npages, DMA_ERROR_CODE);
+	iommu_tbl_range_free(&iommu->tbl, dvma, npages, IOMMU_ERROR_CODE);
 	order = get_order(size);
 	if (order < 10)
 		free_pages((unsigned long)cpu, order);
@@ -256,7 +256,7 @@ static dma_addr_t dma_4v_map_page(struct device *dev, struct page *page,
 	entry = iommu_tbl_range_alloc(dev, &iommu->tbl, npages, NULL,
 				      (unsigned long)(-1), 0);
 
-	if (unlikely(entry == DMA_ERROR_CODE))
+	if (unlikely(entry == IOMMU_ERROR_CODE))
 		goto bad;
 
 	bus_addr = (iommu->tbl.table_map_base + (entry << IO_PAGE_SHIFT));
@@ -288,7 +288,7 @@ bad:
 	return DMA_ERROR_CODE;
 
 iommu_map_fail:
-	iommu_tbl_range_free(&iommu->tbl, bus_addr, npages, DMA_ERROR_CODE);
+	iommu_tbl_range_free(&iommu->tbl, bus_addr, npages, IOMMU_ERROR_CODE);
 	return DMA_ERROR_CODE;
 }
 
@@ -317,7 +317,7 @@ static void dma_4v_unmap_page(struct device *dev, dma_addr_t bus_addr,
 	bus_addr &= IO_PAGE_MASK;
 	entry = (bus_addr - iommu->tbl.table_map_base) >> IO_PAGE_SHIFT;
 	dma_4v_iommu_demap(&devhandle, entry, npages);
-	iommu_tbl_range_free(&iommu->tbl, bus_addr, npages, DMA_ERROR_CODE);
+	iommu_tbl_range_free(&iommu->tbl, bus_addr, npages, IOMMU_ERROR_CODE);
 }
 
 static int dma_4v_map_sg(struct device *dev, struct scatterlist *sglist,
@@ -376,7 +376,7 @@ static int dma_4v_map_sg(struct device *dev, struct scatterlist *sglist,
 					      &handle, (unsigned long)(-1), 0);
 
 		/* Handle failure */
-		if (unlikely(entry == DMA_ERROR_CODE)) {
+		if (unlikely(entry == IOMMU_ERROR_CODE)) {
 			if (printk_ratelimit())
 				printk(KERN_INFO "iommu_alloc failed, iommu %p paddr %lx"
 				       " npages %lx\n", iommu, paddr, npages);
@@ -451,7 +451,7 @@ iommu_map_failed:
 			npages = iommu_num_pages(s->dma_address, s->dma_length,
 						 IO_PAGE_SIZE);
 			iommu_tbl_range_free(&iommu->tbl, vaddr, npages,
-					     DMA_ERROR_CODE);
+					     IOMMU_ERROR_CODE);
 			/* XXX demap? XXX */
 			s->dma_address = DMA_ERROR_CODE;
 			s->dma_length = 0;
@@ -496,7 +496,7 @@ static void dma_4v_unmap_sg(struct device *dev, struct scatterlist *sglist,
 		entry = ((dma_handle - tbl->table_map_base) >> shift);
 		dma_4v_iommu_demap(&devhandle, entry, npages);
 		iommu_tbl_range_free(&iommu->tbl, dma_handle, npages,
-				     DMA_ERROR_CODE);
+				     IOMMU_ERROR_CODE);
 		sg = sg_next(sg);
 	}
 
diff --git a/arch/sparc/kernel/perf_event.c b/arch/sparc/kernel/perf_event.c
index 689db65f8..6596f66ce 100644
--- a/arch/sparc/kernel/perf_event.c
+++ b/arch/sparc/kernel/perf_event.c
@@ -9,7 +9,7 @@
  *  Copyright (C) 2008-2009 Red Hat, Inc., Ingo Molnar
  *  Copyright (C) 2009 Jaswinder Singh Rajput
  *  Copyright (C) 2009 Advanced Micro Devices, Inc., Robert Richter
- *  Copyright (C) 2008-2009 Red Hat, Inc., Peter Zijlstra <pzijlstr@redhat.com>
+ *  Copyright (C) 2008-2009 Red Hat, Inc., Peter Zijlstra
  */
 
 #include <linux/perf_event.h>
@@ -108,7 +108,7 @@ struct cpu_hw_events {
 	/* Enabled/disable state.  */
 	int			enabled;
 
-	unsigned int		group_flag;
+	unsigned int		txn_flags;
 };
 static DEFINE_PER_CPU(struct cpu_hw_events, cpu_hw_events) = { .enabled = 1, };
 
@@ -1379,7 +1379,7 @@ static int sparc_pmu_add(struct perf_event *event, int ef_flags)
 	 * skip the schedulability test here, it will be performed
 	 * at commit time(->commit_txn) as a whole
 	 */
-	if (cpuc->group_flag & PERF_EVENT_TXN)
+	if (cpuc->txn_flags & PERF_PMU_TXN_ADD)
 		goto nocheck;
 
 	if (check_excludes(cpuc->event, n0, 1))
@@ -1494,12 +1494,17 @@ static int sparc_pmu_event_init(struct perf_event *event)
  * Set the flag to make pmu::enable() not perform the
  * schedulability test, it will be performed at commit time
  */
-static void sparc_pmu_start_txn(struct pmu *pmu)
+static void sparc_pmu_start_txn(struct pmu *pmu, unsigned int txn_flags)
 {
 	struct cpu_hw_events *cpuhw = this_cpu_ptr(&cpu_hw_events);
 
+	WARN_ON_ONCE(cpuhw->txn_flags);		/* txn already in flight */
+
+	cpuhw->txn_flags = txn_flags;
+	if (txn_flags & ~PERF_PMU_TXN_ADD)
+		return;
+
 	perf_pmu_disable(pmu);
-	cpuhw->group_flag |= PERF_EVENT_TXN;
 }
 
 /*
@@ -1510,8 +1515,15 @@ static void sparc_pmu_start_txn(struct pmu *pmu)
 static void sparc_pmu_cancel_txn(struct pmu *pmu)
 {
 	struct cpu_hw_events *cpuhw = this_cpu_ptr(&cpu_hw_events);
+	unsigned int txn_flags;
+
+	WARN_ON_ONCE(!cpuhw->txn_flags);	/* no txn in flight */
+
+	txn_flags = cpuhw->txn_flags;
+	cpuhw->txn_flags = 0;
+	if (txn_flags & ~PERF_PMU_TXN_ADD)
+		return;
 
-	cpuhw->group_flag &= ~PERF_EVENT_TXN;
 	perf_pmu_enable(pmu);
 }
 
@@ -1528,14 +1540,20 @@ static int sparc_pmu_commit_txn(struct pmu *pmu)
 	if (!sparc_pmu)
 		return -EINVAL;
 
-	cpuc = this_cpu_ptr(&cpu_hw_events);
+	WARN_ON_ONCE(!cpuc->txn_flags);	/* no txn in flight */
+
+	if (cpuc->txn_flags & ~PERF_PMU_TXN_ADD) {
+		cpuc->txn_flags = 0;
+		return 0;
+	}
+
 	n = cpuc->n_events;
 	if (check_excludes(cpuc->event, 0, n))
 		return -EINVAL;
 	if (sparc_check_constraints(cpuc->event, cpuc->events, n))
 		return -EAGAIN;
 
-	cpuc->group_flag &= ~PERF_EVENT_TXN;
+	cpuc->txn_flags = 0;
 	perf_pmu_enable(pmu);
 	return 0;
 }
@@ -1810,11 +1828,18 @@ static void perf_callchain_user_32(struct perf_callchain_entry *entry,
 void
 perf_callchain_user(struct perf_callchain_entry *entry, struct pt_regs *regs)
 {
+	u64 saved_fault_address = current_thread_info()->fault_address;
+	u8 saved_fault_code = get_thread_fault_code();
+	mm_segment_t old_fs;
+
 	perf_callchain_store(entry, regs->tpc);
 
 	if (!current->mm)
 		return;
 
+	old_fs = get_fs();
+	set_fs(USER_DS);
+
 	flushw_user();
 
 	pagefault_disable();
@@ -1825,4 +1850,8 @@ perf_callchain_user(struct perf_callchain_entry *entry, struct pt_regs *regs)
 		perf_callchain_user_64(entry, regs);
 
 	pagefault_enable();
+
+	set_fs(old_fs);
+	set_thread_fault_code(saved_fault_code);
+	current_thread_info()->fault_address = saved_fault_address;
 }
diff --git a/arch/sparc/kernel/rtrap_64.S b/arch/sparc/kernel/rtrap_64.S
index 39f0c662f..d08bdaffd 100644
--- a/arch/sparc/kernel/rtrap_64.S
+++ b/arch/sparc/kernel/rtrap_64.S
@@ -73,7 +73,13 @@ rtrap_nmi:	ldx			[%sp + PTREGS_OFF + PT_V9_TSTATE], %l1
 		andn			%l1, %l4, %l1
 		srl			%l4, 20, %l4
 		ba,pt			%xcc, rtrap_no_irq_enable
-		 wrpr			%l4, %pil
+		nop
+		/* Do not actually set the %pil here.  We will do that
+		 * below after we clear PSTATE_IE in the %pstate register.
+		 * If we re-enable interrupts here, we can recurse down
+		 * the hardirq stack potentially endlessly, causing a
+		 * stack overflow.
+		 */
 
 		.align			64
 		.globl			rtrap_irq, rtrap, irqsz_patchme, rtrap_xcall
diff --git a/arch/sparc/kernel/setup_64.c b/arch/sparc/kernel/setup_64.c
index f7b261749..f3185e2b0 100644
--- a/arch/sparc/kernel/setup_64.c
+++ b/arch/sparc/kernel/setup_64.c
@@ -380,7 +380,8 @@ static const char *hwcaps[] = {
 	 */
 	"mul32", "div32", "fsmuld", "v8plus", "popc", "vis", "vis2",
 	"ASIBlkInit", "fmaf", "vis3", "hpc", "random", "trans", "fjfmau",
-	"ima", "cspare", "pause", "cbcond",
+	"ima", "cspare", "pause", "cbcond", NULL /*reserved for crypto */,
+	"adp",
 };
 
 static const char *crypto_hwcaps[] = {
@@ -396,7 +397,7 @@ void cpucap_info(struct seq_file *m)
 	seq_puts(m, "cpucaps\t\t: ");
 	for (i = 0; i < ARRAY_SIZE(hwcaps); i++) {
 		unsigned long bit = 1UL << i;
-		if (caps & bit) {
+		if (hwcaps[i] && (caps & bit)) {
 			seq_printf(m, "%s%s",
 				   printed ? "," : "", hwcaps[i]);
 			printed++;
@@ -450,7 +451,7 @@ static void __init report_hwcaps(unsigned long caps)
 
 	for (i = 0; i < ARRAY_SIZE(hwcaps); i++) {
 		unsigned long bit = 1UL << i;
-		if (caps & bit)
+		if (hwcaps[i] && (caps & bit))
 			report_one_hwcap(&printed, hwcaps[i]);
 	}
 	if (caps & HWCAP_SPARC_CRYPTO)
@@ -485,7 +486,7 @@ static unsigned long __init mdesc_cpu_hwcap_list(void)
 		for (i = 0; i < ARRAY_SIZE(hwcaps); i++) {
 			unsigned long bit = 1UL << i;
 
-			if (!strcmp(prop, hwcaps[i])) {
+			if (hwcaps[i] && !strcmp(prop, hwcaps[i])) {
 				caps |= bit;
 				break;
 			}
diff --git a/arch/sparc/kernel/systbls_32.S b/arch/sparc/kernel/systbls_32.S
index e31a9056a..e663b6c78 100644
--- a/arch/sparc/kernel/systbls_32.S
+++ b/arch/sparc/kernel/systbls_32.S
@@ -35,18 +35,18 @@ sys_call_table:
 /*80*/	.long sys_setgroups16, sys_getpgrp, sys_setgroups, sys_setitimer, sys_ftruncate64
 /*85*/	.long sys_swapon, sys_getitimer, sys_setuid, sys_sethostname, sys_setgid
 /*90*/	.long sys_dup2, sys_setfsuid, sys_fcntl, sys_select, sys_setfsgid
-/*95*/	.long sys_fsync, sys_setpriority, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall
+/*95*/	.long sys_fsync, sys_setpriority, sys_socket, sys_connect, sys_accept
 /*100*/	.long sys_getpriority, sys_rt_sigreturn, sys_rt_sigaction, sys_rt_sigprocmask, sys_rt_sigpending
 /*105*/	.long sys_rt_sigtimedwait, sys_rt_sigqueueinfo, sys_rt_sigsuspend, sys_setresuid, sys_getresuid
-/*110*/	.long sys_setresgid, sys_getresgid, sys_setregid, sys_nis_syscall, sys_nis_syscall
-/*115*/	.long sys_getgroups, sys_gettimeofday, sys_getrusage, sys_nis_syscall, sys_getcwd
+/*110*/	.long sys_setresgid, sys_getresgid, sys_setregid, sys_recvmsg, sys_sendmsg
+/*115*/	.long sys_getgroups, sys_gettimeofday, sys_getrusage, sys_getsockopt, sys_getcwd
 /*120*/	.long sys_readv, sys_writev, sys_settimeofday, sys_fchown16, sys_fchmod
-/*125*/	.long sys_nis_syscall, sys_setreuid16, sys_setregid16, sys_rename, sys_truncate
-/*130*/	.long sys_ftruncate, sys_flock, sys_lstat64, sys_nis_syscall, sys_nis_syscall
-/*135*/	.long sys_nis_syscall, sys_mkdir, sys_rmdir, sys_utimes, sys_stat64
-/*140*/	.long sys_sendfile64, sys_nis_syscall, sys_futex, sys_gettid, sys_getrlimit
+/*125*/	.long sys_recvfrom, sys_setreuid16, sys_setregid16, sys_rename, sys_truncate
+/*130*/	.long sys_ftruncate, sys_flock, sys_lstat64, sys_sendto, sys_shutdown
+/*135*/	.long sys_socketpair, sys_mkdir, sys_rmdir, sys_utimes, sys_stat64
+/*140*/	.long sys_sendfile64, sys_getpeername, sys_futex, sys_gettid, sys_getrlimit
 /*145*/	.long sys_setrlimit, sys_pivot_root, sys_prctl, sys_pciconfig_read, sys_pciconfig_write
-/*150*/	.long sys_nis_syscall, sys_inotify_init, sys_inotify_add_watch, sys_poll, sys_getdents64
+/*150*/	.long sys_getsockname, sys_inotify_init, sys_inotify_add_watch, sys_poll, sys_getdents64
 /*155*/	.long sys_fcntl64, sys_inotify_rm_watch, sys_statfs, sys_fstatfs, sys_oldumount
 /*160*/	.long sys_sched_setaffinity, sys_sched_getaffinity, sys_getdomainname, sys_setdomainname, sys_nis_syscall
 /*165*/	.long sys_quotactl, sys_set_tid_address, sys_mount, sys_ustat, sys_setxattr
@@ -87,4 +87,5 @@ sys_call_table:
 /*335*/	.long sys_syncfs, sys_sendmmsg, sys_setns, sys_process_vm_readv, sys_process_vm_writev
 /*340*/	.long sys_ni_syscall, sys_kcmp, sys_finit_module, sys_sched_setattr, sys_sched_getattr
 /*345*/	.long sys_renameat2, sys_seccomp, sys_getrandom, sys_memfd_create, sys_bpf
-/*350*/	.long sys_execveat
+/*350*/	.long sys_execveat, sys_membarrier, sys_userfaultfd, sys_bind, sys_listen
+/*355*/	.long sys_setsockopt, sys_mlock2
diff --git a/arch/sparc/kernel/systbls_64.S b/arch/sparc/kernel/systbls_64.S
index d72f76ae7..1557121f4 100644
--- a/arch/sparc/kernel/systbls_64.S
+++ b/arch/sparc/kernel/systbls_64.S
@@ -37,15 +37,15 @@ sys_call_table32:
 /*80*/	.word sys_setgroups16, sys_getpgrp, sys_setgroups, compat_sys_setitimer, sys32_ftruncate64
 	.word sys_swapon, compat_sys_getitimer, sys_setuid, sys_sethostname, sys_setgid
 /*90*/	.word sys_dup2, sys_setfsuid, compat_sys_fcntl, sys32_select, sys_setfsgid
-	.word sys_fsync, sys_setpriority, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall
+	.word sys_fsync, sys_setpriority, sys_socket, sys_connect, sys_accept
 /*100*/ .word sys_getpriority, sys32_rt_sigreturn, compat_sys_rt_sigaction, compat_sys_rt_sigprocmask, compat_sys_rt_sigpending
 	.word compat_sys_rt_sigtimedwait, compat_sys_rt_sigqueueinfo, compat_sys_rt_sigsuspend, sys_setresuid, sys_getresuid
-/*110*/	.word sys_setresgid, sys_getresgid, sys_setregid, sys_nis_syscall, sys_nis_syscall
-	.word sys_getgroups, compat_sys_gettimeofday, compat_sys_getrusage, sys_nis_syscall, sys_getcwd
+/*110*/	.word sys_setresgid, sys_getresgid, sys_setregid, compat_sys_recvmsg, compat_sys_sendmsg
+	.word sys_getgroups, compat_sys_gettimeofday, compat_sys_getrusage, compat_sys_getsockopt, sys_getcwd
 /*120*/	.word compat_sys_readv, compat_sys_writev, compat_sys_settimeofday, sys_fchown16, sys_fchmod
-	.word sys_nis_syscall, sys_setreuid16, sys_setregid16, sys_rename, compat_sys_truncate
-/*130*/	.word compat_sys_ftruncate, sys_flock, compat_sys_lstat64, sys_nis_syscall, sys_nis_syscall
-	.word sys_nis_syscall, sys_mkdir, sys_rmdir, compat_sys_utimes, compat_sys_stat64
+	.word sys_recvfrom, sys_setreuid16, sys_setregid16, sys_rename, compat_sys_truncate
+/*130*/	.word compat_sys_ftruncate, sys_flock, compat_sys_lstat64, sys_sendto, sys_shutdown
+	.word sys_socketpair, sys_mkdir, sys_rmdir, compat_sys_utimes, compat_sys_stat64
 /*140*/	.word sys_sendfile64, sys_nis_syscall, sys32_futex, sys_gettid, compat_sys_getrlimit
 	.word compat_sys_setrlimit, sys_pivot_root, sys_prctl, sys_pciconfig_read, sys_pciconfig_write
 /*150*/	.word sys_nis_syscall, sys_inotify_init, sys_inotify_add_watch, sys_poll, sys_getdents64
@@ -88,7 +88,8 @@ sys_call_table32:
 	.word sys_syncfs, compat_sys_sendmmsg, sys_setns, compat_sys_process_vm_readv, compat_sys_process_vm_writev
 /*340*/	.word sys_kern_features, sys_kcmp, sys_finit_module, sys_sched_setattr, sys_sched_getattr
 	.word sys32_renameat2, sys_seccomp, sys_getrandom, sys_memfd_create, sys_bpf
-/*350*/	.word sys32_execveat
+/*350*/	.word sys32_execveat, sys_membarrier, sys_userfaultfd, sys_bind, sys_listen
+	.word compat_sys_setsockopt, sys_mlock2
 
 #endif /* CONFIG_COMPAT */
 
@@ -168,4 +169,5 @@ sys_call_table:
 	.word sys_syncfs, sys_sendmmsg, sys_setns, sys_process_vm_readv, sys_process_vm_writev
 /*340*/	.word sys_kern_features, sys_kcmp, sys_finit_module, sys_sched_setattr, sys_sched_getattr
 	.word sys_renameat2, sys_seccomp, sys_getrandom, sys_memfd_create, sys_bpf
-/*350*/	.word sys64_execveat
+/*350*/	.word sys64_execveat, sys_membarrier, sys_userfaultfd, sys_bind, sys_listen
+	.word sys_setsockopt, sys_mlock2
diff --git a/arch/sparc/kernel/unaligned_64.c b/arch/sparc/kernel/unaligned_64.c
index 62098a89b..d89e97b37 100644
--- a/arch/sparc/kernel/unaligned_64.c
+++ b/arch/sparc/kernel/unaligned_64.c
@@ -436,24 +436,26 @@ extern void sun4v_data_access_exception(struct pt_regs *regs,
 int handle_ldf_stq(u32 insn, struct pt_regs *regs)
 {
 	unsigned long addr = compute_effective_address(regs, insn, 0);
-	int freg = ((insn >> 25) & 0x1e) | ((insn >> 20) & 0x20);
+	int freg;
 	struct fpustate *f = FPUSTATE;
 	int asi = decode_asi(insn, regs);
-	int flag = (freg < 32) ? FPRS_DL : FPRS_DU;
+	int flag;
 
 	perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, regs, 0);
 
 	save_and_clear_fpu();
 	current_thread_info()->xfsr[0] &= ~0x1c000;
-	if (freg & 3) {
-		current_thread_info()->xfsr[0] |= (6 << 14) /* invalid_fp_register */;
-		do_fpother(regs);
-		return 0;
-	}
 	if (insn & 0x200000) {
 		/* STQ */
 		u64 first = 0, second = 0;
 		
+		freg = ((insn >> 25) & 0x1e) | ((insn >> 20) & 0x20);
+		flag = (freg < 32) ? FPRS_DL : FPRS_DU;
+		if (freg & 3) {
+			current_thread_info()->xfsr[0] |= (6 << 14) /* invalid_fp_register */;
+			do_fpother(regs);
+			return 0;
+		}
 		if (current_thread_info()->fpsaved[0] & flag) {
 			first = *(u64 *)&f->regs[freg];
 			second = *(u64 *)&f->regs[freg+2];
@@ -513,6 +515,12 @@ int handle_ldf_stq(u32 insn, struct pt_regs *regs)
 		case 0x100000: size = 4; break;
 		default: size = 2; break;
 		}
+		if (size == 1)
+			freg = (insn >> 25) & 0x1f;
+		else
+			freg = ((insn >> 25) & 0x1e) | ((insn >> 20) & 0x20);
+		flag = (freg < 32) ? FPRS_DL : FPRS_DU;
+
 		for (i = 0; i < size; i++)
 			data[i] = 0;
 		
diff --git a/arch/sparc/lib/NG2copy_from_user.S b/arch/sparc/lib/NG2copy_from_user.S
index 119ccb9a5..d5242b8c4 100644
--- a/arch/sparc/lib/NG2copy_from_user.S
+++ b/arch/sparc/lib/NG2copy_from_user.S
@@ -11,6 +11,14 @@
 	.text;			\
 	.align 4;
 
+#define EX_LD_FP(x)		\
+98:	x;			\
+	.section __ex_table,"a";\
+	.align 4;		\
+	.word 98b, __retl_one_asi_fp;\
+	.text;			\
+	.align 4;
+
 #ifndef ASI_AIUS
 #define ASI_AIUS	0x11
 #endif
diff --git a/arch/sparc/lib/NG2copy_to_user.S b/arch/sparc/lib/NG2copy_to_user.S
index 7fe1ccefd..4e962d993 100644
--- a/arch/sparc/lib/NG2copy_to_user.S
+++ b/arch/sparc/lib/NG2copy_to_user.S
@@ -11,6 +11,14 @@
 	.text;			\
 	.align 4;
 
+#define EX_ST_FP(x)		\
+98:	x;			\
+	.section __ex_table,"a";\
+	.align 4;		\
+	.word 98b, __retl_one_asi_fp;\
+	.text;			\
+	.align 4;
+
 #ifndef ASI_AIUS
 #define ASI_AIUS	0x11
 #endif
diff --git a/arch/sparc/lib/NG2memcpy.S b/arch/sparc/lib/NG2memcpy.S
index 30eee6e8a..d5f585df2 100644
--- a/arch/sparc/lib/NG2memcpy.S
+++ b/arch/sparc/lib/NG2memcpy.S
@@ -34,10 +34,16 @@
 #ifndef EX_LD
 #define EX_LD(x)	x
 #endif
+#ifndef EX_LD_FP
+#define EX_LD_FP(x)	x
+#endif
 
 #ifndef EX_ST
 #define EX_ST(x)	x
 #endif
+#ifndef EX_ST_FP
+#define EX_ST_FP(x)	x
+#endif
 
 #ifndef EX_RETVAL
 #define EX_RETVAL(x)	x
@@ -134,40 +140,40 @@
 	fsrc2		%x6, %f12; \
 	fsrc2		%x7, %f14;
 #define FREG_LOAD_1(base, x0) \
-	EX_LD(LOAD(ldd, base + 0x00, %x0))
+	EX_LD_FP(LOAD(ldd, base + 0x00, %x0))
 #define FREG_LOAD_2(base, x0, x1) \
-	EX_LD(LOAD(ldd, base + 0x00, %x0)); \
-	EX_LD(LOAD(ldd, base + 0x08, %x1));
+	EX_LD_FP(LOAD(ldd, base + 0x00, %x0)); \
+	EX_LD_FP(LOAD(ldd, base + 0x08, %x1));
 #define FREG_LOAD_3(base, x0, x1, x2) \
-	EX_LD(LOAD(ldd, base + 0x00, %x0)); \
-	EX_LD(LOAD(ldd, base + 0x08, %x1)); \
-	EX_LD(LOAD(ldd, base + 0x10, %x2));
+	EX_LD_FP(LOAD(ldd, base + 0x00, %x0)); \
+	EX_LD_FP(LOAD(ldd, base + 0x08, %x1)); \
+	EX_LD_FP(LOAD(ldd, base + 0x10, %x2));
 #define FREG_LOAD_4(base, x0, x1, x2, x3) \
-	EX_LD(LOAD(ldd, base + 0x00, %x0)); \
-	EX_LD(LOAD(ldd, base + 0x08, %x1)); \
-	EX_LD(LOAD(ldd, base + 0x10, %x2)); \
-	EX_LD(LOAD(ldd, base + 0x18, %x3));
+	EX_LD_FP(LOAD(ldd, base + 0x00, %x0)); \
+	EX_LD_FP(LOAD(ldd, base + 0x08, %x1)); \
+	EX_LD_FP(LOAD(ldd, base + 0x10, %x2)); \
+	EX_LD_FP(LOAD(ldd, base + 0x18, %x3));
 #define FREG_LOAD_5(base, x0, x1, x2, x3, x4) \
-	EX_LD(LOAD(ldd, base + 0x00, %x0)); \
-	EX_LD(LOAD(ldd, base + 0x08, %x1)); \
-	EX_LD(LOAD(ldd, base + 0x10, %x2)); \
-	EX_LD(LOAD(ldd, base + 0x18, %x3)); \
-	EX_LD(LOAD(ldd, base + 0x20, %x4));
+	EX_LD_FP(LOAD(ldd, base + 0x00, %x0)); \
+	EX_LD_FP(LOAD(ldd, base + 0x08, %x1)); \
+	EX_LD_FP(LOAD(ldd, base + 0x10, %x2)); \
+	EX_LD_FP(LOAD(ldd, base + 0x18, %x3)); \
+	EX_LD_FP(LOAD(ldd, base + 0x20, %x4));
 #define FREG_LOAD_6(base, x0, x1, x2, x3, x4, x5) \
-	EX_LD(LOAD(ldd, base + 0x00, %x0)); \
-	EX_LD(LOAD(ldd, base + 0x08, %x1)); \
-	EX_LD(LOAD(ldd, base + 0x10, %x2)); \
-	EX_LD(LOAD(ldd, base + 0x18, %x3)); \
-	EX_LD(LOAD(ldd, base + 0x20, %x4)); \
-	EX_LD(LOAD(ldd, base + 0x28, %x5));
+	EX_LD_FP(LOAD(ldd, base + 0x00, %x0)); \
+	EX_LD_FP(LOAD(ldd, base + 0x08, %x1)); \
+	EX_LD_FP(LOAD(ldd, base + 0x10, %x2)); \
+	EX_LD_FP(LOAD(ldd, base + 0x18, %x3)); \
+	EX_LD_FP(LOAD(ldd, base + 0x20, %x4)); \
+	EX_LD_FP(LOAD(ldd, base + 0x28, %x5));
 #define FREG_LOAD_7(base, x0, x1, x2, x3, x4, x5, x6) \
-	EX_LD(LOAD(ldd, base + 0x00, %x0)); \
-	EX_LD(LOAD(ldd, base + 0x08, %x1)); \
-	EX_LD(LOAD(ldd, base + 0x10, %x2)); \
-	EX_LD(LOAD(ldd, base + 0x18, %x3)); \
-	EX_LD(LOAD(ldd, base + 0x20, %x4)); \
-	EX_LD(LOAD(ldd, base + 0x28, %x5)); \
-	EX_LD(LOAD(ldd, base + 0x30, %x6));
+	EX_LD_FP(LOAD(ldd, base + 0x00, %x0)); \
+	EX_LD_FP(LOAD(ldd, base + 0x08, %x1)); \
+	EX_LD_FP(LOAD(ldd, base + 0x10, %x2)); \
+	EX_LD_FP(LOAD(ldd, base + 0x18, %x3)); \
+	EX_LD_FP(LOAD(ldd, base + 0x20, %x4)); \
+	EX_LD_FP(LOAD(ldd, base + 0x28, %x5)); \
+	EX_LD_FP(LOAD(ldd, base + 0x30, %x6));
 
 	.register	%g2,#scratch
 	.register	%g3,#scratch
@@ -275,11 +281,11 @@ FUNC_NAME:	/* %o0=dst, %o1=src, %o2=len */
 	 nop
 	/* fall through for 0 < low bits < 8 */
 110:	sub		%o4, 64, %g2
-	EX_LD(LOAD_BLK(%g2, %f0))
-1:	EX_ST(STORE_INIT(%g0, %o4 + %g3))
-	EX_LD(LOAD_BLK(%o4, %f16))
+	EX_LD_FP(LOAD_BLK(%g2, %f0))
+1:	EX_ST_FP(STORE_INIT(%g0, %o4 + %g3))
+	EX_LD_FP(LOAD_BLK(%o4, %f16))
 	FREG_FROB(f0, f2, f4, f6, f8, f10, f12, f14, f16)
-	EX_ST(STORE_BLK(%f0, %o4 + %g3))
+	EX_ST_FP(STORE_BLK(%f0, %o4 + %g3))
 	FREG_MOVE_8(f16, f18, f20, f22, f24, f26, f28, f30)
 	subcc		%g1, 64, %g1
 	add		%o4, 64, %o4
@@ -290,10 +296,10 @@ FUNC_NAME:	/* %o0=dst, %o1=src, %o2=len */
 
 120:	sub		%o4, 56, %g2
 	FREG_LOAD_7(%g2, f0, f2, f4, f6, f8, f10, f12)
-1:	EX_ST(STORE_INIT(%g0, %o4 + %g3))
-	EX_LD(LOAD_BLK(%o4, %f16))
+1:	EX_ST_FP(STORE_INIT(%g0, %o4 + %g3))
+	EX_LD_FP(LOAD_BLK(%o4, %f16))
 	FREG_FROB(f0, f2, f4, f6, f8, f10, f12, f16, f18)
-	EX_ST(STORE_BLK(%f0, %o4 + %g3))
+	EX_ST_FP(STORE_BLK(%f0, %o4 + %g3))
 	FREG_MOVE_7(f18, f20, f22, f24, f26, f28, f30)
 	subcc		%g1, 64, %g1
 	add		%o4, 64, %o4
@@ -304,10 +310,10 @@ FUNC_NAME:	/* %o0=dst, %o1=src, %o2=len */
 
 130:	sub		%o4, 48, %g2
 	FREG_LOAD_6(%g2, f0, f2, f4, f6, f8, f10)
-1:	EX_ST(STORE_INIT(%g0, %o4 + %g3))
-	EX_LD(LOAD_BLK(%o4, %f16))
+1:	EX_ST_FP(STORE_INIT(%g0, %o4 + %g3))
+	EX_LD_FP(LOAD_BLK(%o4, %f16))
 	FREG_FROB(f0, f2, f4, f6, f8, f10, f16, f18, f20)
-	EX_ST(STORE_BLK(%f0, %o4 + %g3))
+	EX_ST_FP(STORE_BLK(%f0, %o4 + %g3))
 	FREG_MOVE_6(f20, f22, f24, f26, f28, f30)
 	subcc		%g1, 64, %g1
 	add		%o4, 64, %o4
@@ -318,10 +324,10 @@ FUNC_NAME:	/* %o0=dst, %o1=src, %o2=len */
 
 140:	sub		%o4, 40, %g2
 	FREG_LOAD_5(%g2, f0, f2, f4, f6, f8)
-1:	EX_ST(STORE_INIT(%g0, %o4 + %g3))
-	EX_LD(LOAD_BLK(%o4, %f16))
+1:	EX_ST_FP(STORE_INIT(%g0, %o4 + %g3))
+	EX_LD_FP(LOAD_BLK(%o4, %f16))
 	FREG_FROB(f0, f2, f4, f6, f8, f16, f18, f20, f22)
-	EX_ST(STORE_BLK(%f0, %o4 + %g3))
+	EX_ST_FP(STORE_BLK(%f0, %o4 + %g3))
 	FREG_MOVE_5(f22, f24, f26, f28, f30)
 	subcc		%g1, 64, %g1
 	add		%o4, 64, %o4
@@ -332,10 +338,10 @@ FUNC_NAME:	/* %o0=dst, %o1=src, %o2=len */
 
 150:	sub		%o4, 32, %g2
 	FREG_LOAD_4(%g2, f0, f2, f4, f6)
-1:	EX_ST(STORE_INIT(%g0, %o4 + %g3))
-	EX_LD(LOAD_BLK(%o4, %f16))
+1:	EX_ST_FP(STORE_INIT(%g0, %o4 + %g3))
+	EX_LD_FP(LOAD_BLK(%o4, %f16))
 	FREG_FROB(f0, f2, f4, f6, f16, f18, f20, f22, f24)
-	EX_ST(STORE_BLK(%f0, %o4 + %g3))
+	EX_ST_FP(STORE_BLK(%f0, %o4 + %g3))
 	FREG_MOVE_4(f24, f26, f28, f30)
 	subcc		%g1, 64, %g1
 	add		%o4, 64, %o4
@@ -346,10 +352,10 @@ FUNC_NAME:	/* %o0=dst, %o1=src, %o2=len */
 
 160:	sub		%o4, 24, %g2
 	FREG_LOAD_3(%g2, f0, f2, f4)
-1:	EX_ST(STORE_INIT(%g0, %o4 + %g3))
-	EX_LD(LOAD_BLK(%o4, %f16))
+1:	EX_ST_FP(STORE_INIT(%g0, %o4 + %g3))
+	EX_LD_FP(LOAD_BLK(%o4, %f16))
 	FREG_FROB(f0, f2, f4, f16, f18, f20, f22, f24, f26)
-	EX_ST(STORE_BLK(%f0, %o4 + %g3))
+	EX_ST_FP(STORE_BLK(%f0, %o4 + %g3))
 	FREG_MOVE_3(f26, f28, f30)
 	subcc		%g1, 64, %g1
 	add		%o4, 64, %o4
@@ -360,10 +366,10 @@ FUNC_NAME:	/* %o0=dst, %o1=src, %o2=len */
 
 170:	sub		%o4, 16, %g2
 	FREG_LOAD_2(%g2, f0, f2)
-1:	EX_ST(STORE_INIT(%g0, %o4 + %g3))
-	EX_LD(LOAD_BLK(%o4, %f16))
+1:	EX_ST_FP(STORE_INIT(%g0, %o4 + %g3))
+	EX_LD_FP(LOAD_BLK(%o4, %f16))
 	FREG_FROB(f0, f2, f16, f18, f20, f22, f24, f26, f28)
-	EX_ST(STORE_BLK(%f0, %o4 + %g3))
+	EX_ST_FP(STORE_BLK(%f0, %o4 + %g3))
 	FREG_MOVE_2(f28, f30)
 	subcc		%g1, 64, %g1
 	add		%o4, 64, %o4
@@ -374,10 +380,10 @@ FUNC_NAME:	/* %o0=dst, %o1=src, %o2=len */
 
 180:	sub		%o4, 8, %g2
 	FREG_LOAD_1(%g2, f0)
-1:	EX_ST(STORE_INIT(%g0, %o4 + %g3))
-	EX_LD(LOAD_BLK(%o4, %f16))
+1:	EX_ST_FP(STORE_INIT(%g0, %o4 + %g3))
+	EX_LD_FP(LOAD_BLK(%o4, %f16))
 	FREG_FROB(f0, f16, f18, f20, f22, f24, f26, f28, f30)
-	EX_ST(STORE_BLK(%f0, %o4 + %g3))
+	EX_ST_FP(STORE_BLK(%f0, %o4 + %g3))
 	FREG_MOVE_1(f30)
 	subcc		%g1, 64, %g1
 	add		%o4, 64, %o4
@@ -387,10 +393,10 @@ FUNC_NAME:	/* %o0=dst, %o1=src, %o2=len */
 	 nop
 
 190:
-1:	EX_ST(STORE_INIT(%g0, %o4 + %g3))
+1:	EX_ST_FP(STORE_INIT(%g0, %o4 + %g3))
 	subcc		%g1, 64, %g1
-	EX_LD(LOAD_BLK(%o4, %f0))
-	EX_ST(STORE_BLK(%f0, %o4 + %g3))
+	EX_LD_FP(LOAD_BLK(%o4, %f0))
+	EX_ST_FP(STORE_BLK(%f0, %o4 + %g3))
 	add		%o4, 64, %o4
 	bne,pt		%xcc, 1b
 	 LOAD(prefetch, %o4 + 64, #one_read)
diff --git a/arch/sparc/lib/NG4copy_from_user.S b/arch/sparc/lib/NG4copy_from_user.S
index fd9f903ff..2e8ee7ad0 100644
--- a/arch/sparc/lib/NG4copy_from_user.S
+++ b/arch/sparc/lib/NG4copy_from_user.S
@@ -11,6 +11,14 @@
 	.text;			\
 	.align 4;
 
+#define EX_LD_FP(x)		\
+98:	x;			\
+	.section __ex_table,"a";\
+	.align 4;		\
+	.word 98b, __retl_one_asi_fp;\
+	.text;			\
+	.align 4;
+
 #ifndef ASI_AIUS
 #define ASI_AIUS	0x11
 #endif
diff --git a/arch/sparc/lib/NG4copy_to_user.S b/arch/sparc/lib/NG4copy_to_user.S
index 9744c4540..be0bf4590 100644
--- a/arch/sparc/lib/NG4copy_to_user.S
+++ b/arch/sparc/lib/NG4copy_to_user.S
@@ -11,6 +11,14 @@
 	.text;			\
 	.align 4;
 
+#define EX_ST_FP(x)		\
+98:	x;			\
+	.section __ex_table,"a";\
+	.align 4;		\
+	.word 98b, __retl_one_asi_fp;\
+	.text;			\
+	.align 4;
+
 #ifndef ASI_AIUS
 #define ASI_AIUS	0x11
 #endif
diff --git a/arch/sparc/lib/NG4memcpy.S b/arch/sparc/lib/NG4memcpy.S
index 83aeeb1df..8e13ee1f4 100644
--- a/arch/sparc/lib/NG4memcpy.S
+++ b/arch/sparc/lib/NG4memcpy.S
@@ -48,10 +48,16 @@
 #ifndef EX_LD
 #define EX_LD(x)	x
 #endif
+#ifndef EX_LD_FP
+#define EX_LD_FP(x)	x
+#endif
 
 #ifndef EX_ST
 #define EX_ST(x)	x
 #endif
+#ifndef EX_ST_FP
+#define EX_ST_FP(x)	x
+#endif
 
 #ifndef EX_RETVAL
 #define EX_RETVAL(x)	x
@@ -210,17 +216,17 @@ FUNC_NAME:	/* %o0=dst, %o1=src, %o2=len */
 	sub		%o2, %o4, %o2
 	alignaddr	%o1, %g0, %g1
 	add		%o1, %o4, %o1
-	EX_LD(LOAD(ldd, %g1 + 0x00, %f0))
-1:	EX_LD(LOAD(ldd, %g1 + 0x08, %f2))
+	EX_LD_FP(LOAD(ldd, %g1 + 0x00, %f0))
+1:	EX_LD_FP(LOAD(ldd, %g1 + 0x08, %f2))
 	subcc		%o4, 0x40, %o4
-	EX_LD(LOAD(ldd, %g1 + 0x10, %f4))
-	EX_LD(LOAD(ldd, %g1 + 0x18, %f6))
-	EX_LD(LOAD(ldd, %g1 + 0x20, %f8))
-	EX_LD(LOAD(ldd, %g1 + 0x28, %f10))
-	EX_LD(LOAD(ldd, %g1 + 0x30, %f12))
-	EX_LD(LOAD(ldd, %g1 + 0x38, %f14))
+	EX_LD_FP(LOAD(ldd, %g1 + 0x10, %f4))
+	EX_LD_FP(LOAD(ldd, %g1 + 0x18, %f6))
+	EX_LD_FP(LOAD(ldd, %g1 + 0x20, %f8))
+	EX_LD_FP(LOAD(ldd, %g1 + 0x28, %f10))
+	EX_LD_FP(LOAD(ldd, %g1 + 0x30, %f12))
+	EX_LD_FP(LOAD(ldd, %g1 + 0x38, %f14))
 	faligndata	%f0, %f2, %f16
-	EX_LD(LOAD(ldd, %g1 + 0x40, %f0))
+	EX_LD_FP(LOAD(ldd, %g1 + 0x40, %f0))
 	faligndata	%f2, %f4, %f18
 	add		%g1, 0x40, %g1
 	faligndata	%f4, %f6, %f20
@@ -229,14 +235,14 @@ FUNC_NAME:	/* %o0=dst, %o1=src, %o2=len */
 	faligndata	%f10, %f12, %f26
 	faligndata	%f12, %f14, %f28
 	faligndata	%f14, %f0, %f30
-	EX_ST(STORE(std, %f16, %o0 + 0x00))
-	EX_ST(STORE(std, %f18, %o0 + 0x08))
-	EX_ST(STORE(std, %f20, %o0 + 0x10))
-	EX_ST(STORE(std, %f22, %o0 + 0x18))
-	EX_ST(STORE(std, %f24, %o0 + 0x20))
-	EX_ST(STORE(std, %f26, %o0 + 0x28))
-	EX_ST(STORE(std, %f28, %o0 + 0x30))
-	EX_ST(STORE(std, %f30, %o0 + 0x38))
+	EX_ST_FP(STORE(std, %f16, %o0 + 0x00))
+	EX_ST_FP(STORE(std, %f18, %o0 + 0x08))
+	EX_ST_FP(STORE(std, %f20, %o0 + 0x10))
+	EX_ST_FP(STORE(std, %f22, %o0 + 0x18))
+	EX_ST_FP(STORE(std, %f24, %o0 + 0x20))
+	EX_ST_FP(STORE(std, %f26, %o0 + 0x28))
+	EX_ST_FP(STORE(std, %f28, %o0 + 0x30))
+	EX_ST_FP(STORE(std, %f30, %o0 + 0x38))
 	add		%o0, 0x40, %o0
 	bne,pt		%icc, 1b
 	 LOAD(prefetch, %g1 + 0x200, #n_reads_strong)
diff --git a/arch/sparc/lib/U1copy_from_user.S b/arch/sparc/lib/U1copy_from_user.S
index a6ae2ea04..ecc5692fa 100644
--- a/arch/sparc/lib/U1copy_from_user.S
+++ b/arch/sparc/lib/U1copy_from_user.S
@@ -11,6 +11,14 @@
 	.text;			\
 	.align 4;
 
+#define EX_LD_FP(x)		\
+98:	x;			\
+	.section __ex_table,"a";\
+	.align 4;		\
+	.word 98b, __retl_one_fp;\
+	.text;			\
+	.align 4;
+
 #define FUNC_NAME		___copy_from_user
 #define LOAD(type,addr,dest)	type##a [addr] %asi, dest
 #define LOAD_BLK(addr,dest)	ldda [addr] ASI_BLK_AIUS, dest
diff --git a/arch/sparc/lib/U1copy_to_user.S b/arch/sparc/lib/U1copy_to_user.S
index f4b970eeb..9eea392e4 100644
--- a/arch/sparc/lib/U1copy_to_user.S
+++ b/arch/sparc/lib/U1copy_to_user.S
@@ -11,6 +11,14 @@
 	.text;			\
 	.align 4;
 
+#define EX_ST_FP(x)		\
+98:	x;			\
+	.section __ex_table,"a";\
+	.align 4;		\
+	.word 98b, __retl_one_fp;\
+	.text;			\
+	.align 4;
+
 #define FUNC_NAME		___copy_to_user
 #define STORE(type,src,addr)	type##a src, [addr] ASI_AIUS
 #define STORE_BLK(src,addr)	stda src, [addr] ASI_BLK_AIUS
diff --git a/arch/sparc/lib/U1memcpy.S b/arch/sparc/lib/U1memcpy.S
index b67142b77..3e6209ebb 100644
--- a/arch/sparc/lib/U1memcpy.S
+++ b/arch/sparc/lib/U1memcpy.S
@@ -25,10 +25,16 @@
 #ifndef EX_LD
 #define EX_LD(x)	x
 #endif
+#ifndef EX_LD_FP
+#define EX_LD_FP(x)	x
+#endif
 
 #ifndef EX_ST
 #define EX_ST(x)	x
 #endif
+#ifndef EX_ST_FP
+#define EX_ST_FP(x)	x
+#endif
 
 #ifndef EX_RETVAL
 #define EX_RETVAL(x)	x
@@ -73,8 +79,8 @@
 	faligndata		%f8, %f9, %f62;
 
 #define MAIN_LOOP_CHUNK(src, dest, fdest, fsrc, len, jmptgt)	\
-	EX_LD(LOAD_BLK(%src, %fdest));				\
-	EX_ST(STORE_BLK(%fsrc, %dest));				\
+	EX_LD_FP(LOAD_BLK(%src, %fdest));				\
+	EX_ST_FP(STORE_BLK(%fsrc, %dest));				\
 	add			%src, 0x40, %src;		\
 	subcc			%len, 0x40, %len;		\
 	be,pn			%xcc, jmptgt;			\
@@ -89,12 +95,12 @@
 
 #define DO_SYNC			membar	#Sync;
 #define STORE_SYNC(dest, fsrc)				\
-	EX_ST(STORE_BLK(%fsrc, %dest));			\
+	EX_ST_FP(STORE_BLK(%fsrc, %dest));			\
 	add			%dest, 0x40, %dest;	\
 	DO_SYNC
 
 #define STORE_JUMP(dest, fsrc, target)			\
-	EX_ST(STORE_BLK(%fsrc, %dest));			\
+	EX_ST_FP(STORE_BLK(%fsrc, %dest));			\
 	add			%dest, 0x40, %dest;	\
 	ba,pt			%xcc, target;		\
 	 nop;
@@ -103,7 +109,7 @@
 	subcc			%left, 8, %left;\
 	bl,pn			%xcc, 95f;	\
 	 faligndata		%f0, %f1, %f48;	\
-	EX_ST(STORE(std, %f48, %dest));		\
+	EX_ST_FP(STORE(std, %f48, %dest));		\
 	add			%dest, 8, %dest;
 
 #define UNEVEN_VISCHUNK_LAST(dest, f0, f1, left)	\
@@ -160,8 +166,8 @@ FUNC_NAME:		/* %o0=dst, %o1=src, %o2=len */
 	 and		%g2, 0x38, %g2
 
 1:	subcc		%g1, 0x1, %g1
-	EX_LD(LOAD(ldub, %o1 + 0x00, %o3))
-	EX_ST(STORE(stb, %o3, %o1 + %GLOBAL_SPARE))
+	EX_LD_FP(LOAD(ldub, %o1 + 0x00, %o3))
+	EX_ST_FP(STORE(stb, %o3, %o1 + %GLOBAL_SPARE))
 	bgu,pt		%XCC, 1b
 	 add		%o1, 0x1, %o1
 
@@ -172,20 +178,20 @@ FUNC_NAME:		/* %o0=dst, %o1=src, %o2=len */
 	be,pt		%icc, 3f
 	 alignaddr	%o1, %g0, %o1
 
-	EX_LD(LOAD(ldd, %o1, %f4))
-1:	EX_LD(LOAD(ldd, %o1 + 0x8, %f6))
+	EX_LD_FP(LOAD(ldd, %o1, %f4))
+1:	EX_LD_FP(LOAD(ldd, %o1 + 0x8, %f6))
 	add		%o1, 0x8, %o1
 	subcc		%g2, 0x8, %g2
 	faligndata	%f4, %f6, %f0
-	EX_ST(STORE(std, %f0, %o0))
+	EX_ST_FP(STORE(std, %f0, %o0))
 	be,pn		%icc, 3f
 	 add		%o0, 0x8, %o0
 
-	EX_LD(LOAD(ldd, %o1 + 0x8, %f4))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x8, %f4))
 	add		%o1, 0x8, %o1
 	subcc		%g2, 0x8, %g2
 	faligndata	%f6, %f4, %f0
-	EX_ST(STORE(std, %f0, %o0))
+	EX_ST_FP(STORE(std, %f0, %o0))
 	bne,pt		%icc, 1b
 	 add		%o0, 0x8, %o0
 
@@ -208,13 +214,13 @@ FUNC_NAME:		/* %o0=dst, %o1=src, %o2=len */
 	add		%g1, %GLOBAL_SPARE, %g1
 	subcc		%o2, %g3, %o2
 
-	EX_LD(LOAD_BLK(%o1, %f0))
+	EX_LD_FP(LOAD_BLK(%o1, %f0))
 	add		%o1, 0x40, %o1
 	add		%g1, %g3, %g1
-	EX_LD(LOAD_BLK(%o1, %f16))
+	EX_LD_FP(LOAD_BLK(%o1, %f16))
 	add		%o1, 0x40, %o1
 	sub		%GLOBAL_SPARE, 0x80, %GLOBAL_SPARE
-	EX_LD(LOAD_BLK(%o1, %f32))
+	EX_LD_FP(LOAD_BLK(%o1, %f32))
 	add		%o1, 0x40, %o1
 
 	/* There are 8 instances of the unrolled loop,
@@ -426,28 +432,28 @@ FUNC_NAME:		/* %o0=dst, %o1=src, %o2=len */
 62:	FINISH_VISCHUNK(o0, f44, f46, g3)
 63:	UNEVEN_VISCHUNK_LAST(o0, f46, f0,  g3)
 
-93:	EX_LD(LOAD(ldd, %o1, %f2))
+93:	EX_LD_FP(LOAD(ldd, %o1, %f2))
 	add		%o1, 8, %o1
 	subcc		%g3, 8, %g3
 	faligndata	%f0, %f2, %f8
-	EX_ST(STORE(std, %f8, %o0))
+	EX_ST_FP(STORE(std, %f8, %o0))
 	bl,pn		%xcc, 95f
 	 add		%o0, 8, %o0
-	EX_LD(LOAD(ldd, %o1, %f0))
+	EX_LD_FP(LOAD(ldd, %o1, %f0))
 	add		%o1, 8, %o1
 	subcc		%g3, 8, %g3
 	faligndata	%f2, %f0, %f8
-	EX_ST(STORE(std, %f8, %o0))
+	EX_ST_FP(STORE(std, %f8, %o0))
 	bge,pt		%xcc, 93b
 	 add		%o0, 8, %o0
 
 95:	brz,pt		%o2, 2f
 	 mov		%g1, %o1
 
-1:	EX_LD(LOAD(ldub, %o1, %o3))
+1:	EX_LD_FP(LOAD(ldub, %o1, %o3))
 	add		%o1, 1, %o1
 	subcc		%o2, 1, %o2
-	EX_ST(STORE(stb, %o3, %o0))
+	EX_ST_FP(STORE(stb, %o3, %o0))
 	bne,pt		%xcc, 1b
 	 add		%o0, 1, %o0
 
diff --git a/arch/sparc/lib/U3copy_from_user.S b/arch/sparc/lib/U3copy_from_user.S
index b1acd1331..88ad73d86 100644
--- a/arch/sparc/lib/U3copy_from_user.S
+++ b/arch/sparc/lib/U3copy_from_user.S
@@ -11,6 +11,14 @@
 	.text;			\
 	.align 4;
 
+#define EX_LD_FP(x)		\
+98:	x;			\
+	.section __ex_table,"a";\
+	.align 4;		\
+	.word 98b, __retl_one_fp;\
+	.text;			\
+	.align 4;
+
 #define FUNC_NAME		U3copy_from_user
 #define LOAD(type,addr,dest)	type##a [addr] %asi, dest
 #define EX_RETVAL(x)		0
diff --git a/arch/sparc/lib/U3copy_to_user.S b/arch/sparc/lib/U3copy_to_user.S
index ef1e493af..845139d75 100644
--- a/arch/sparc/lib/U3copy_to_user.S
+++ b/arch/sparc/lib/U3copy_to_user.S
@@ -11,6 +11,14 @@
 	.text;			\
 	.align 4;
 
+#define EX_ST_FP(x)		\
+98:	x;			\
+	.section __ex_table,"a";\
+	.align 4;		\
+	.word 98b, __retl_one_fp;\
+	.text;			\
+	.align 4;
+
 #define FUNC_NAME		U3copy_to_user
 #define STORE(type,src,addr)	type##a src, [addr] ASI_AIUS
 #define STORE_BLK(src,addr)	stda src, [addr] ASI_BLK_AIUS
diff --git a/arch/sparc/lib/U3memcpy.S b/arch/sparc/lib/U3memcpy.S
index 7cae9cc6a..491ee69e4 100644
--- a/arch/sparc/lib/U3memcpy.S
+++ b/arch/sparc/lib/U3memcpy.S
@@ -24,10 +24,16 @@
 #ifndef EX_LD
 #define EX_LD(x)	x
 #endif
+#ifndef EX_LD_FP
+#define EX_LD_FP(x)	x
+#endif
 
 #ifndef EX_ST
 #define EX_ST(x)	x
 #endif
+#ifndef EX_ST_FP
+#define EX_ST_FP(x)	x
+#endif
 
 #ifndef EX_RETVAL
 #define EX_RETVAL(x)	x
@@ -120,8 +126,8 @@ FUNC_NAME:	/* %o0=dst, %o1=src, %o2=len */
 	 and		%g2, 0x38, %g2
 
 1:	subcc		%g1, 0x1, %g1
-	EX_LD(LOAD(ldub, %o1 + 0x00, %o3))
-	EX_ST(STORE(stb, %o3, %o1 + GLOBAL_SPARE))
+	EX_LD_FP(LOAD(ldub, %o1 + 0x00, %o3))
+	EX_ST_FP(STORE(stb, %o3, %o1 + GLOBAL_SPARE))
 	bgu,pt		%XCC, 1b
 	 add		%o1, 0x1, %o1
 
@@ -132,20 +138,20 @@ FUNC_NAME:	/* %o0=dst, %o1=src, %o2=len */
 	be,pt		%icc, 3f
 	 alignaddr	%o1, %g0, %o1
 
-	EX_LD(LOAD(ldd, %o1, %f4))
-1:	EX_LD(LOAD(ldd, %o1 + 0x8, %f6))
+	EX_LD_FP(LOAD(ldd, %o1, %f4))
+1:	EX_LD_FP(LOAD(ldd, %o1 + 0x8, %f6))
 	add		%o1, 0x8, %o1
 	subcc		%g2, 0x8, %g2
 	faligndata	%f4, %f6, %f0
-	EX_ST(STORE(std, %f0, %o0))
+	EX_ST_FP(STORE(std, %f0, %o0))
 	be,pn		%icc, 3f
 	 add		%o0, 0x8, %o0
 
-	EX_LD(LOAD(ldd, %o1 + 0x8, %f4))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x8, %f4))
 	add		%o1, 0x8, %o1
 	subcc		%g2, 0x8, %g2
 	faligndata	%f6, %f4, %f2
-	EX_ST(STORE(std, %f2, %o0))
+	EX_ST_FP(STORE(std, %f2, %o0))
 	bne,pt		%icc, 1b
 	 add		%o0, 0x8, %o0
 
@@ -155,25 +161,25 @@ FUNC_NAME:	/* %o0=dst, %o1=src, %o2=len */
 	LOAD(prefetch, %o1 + 0x080, #one_read)
 	LOAD(prefetch, %o1 + 0x0c0, #one_read)
 	LOAD(prefetch, %o1 + 0x100, #one_read)
-	EX_LD(LOAD(ldd, %o1 + 0x000, %f0))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x000, %f0))
 	LOAD(prefetch, %o1 + 0x140, #one_read)
-	EX_LD(LOAD(ldd, %o1 + 0x008, %f2))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x008, %f2))
 	LOAD(prefetch, %o1 + 0x180, #one_read)
-	EX_LD(LOAD(ldd, %o1 + 0x010, %f4))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x010, %f4))
 	LOAD(prefetch, %o1 + 0x1c0, #one_read)
 	faligndata	%f0, %f2, %f16
-	EX_LD(LOAD(ldd, %o1 + 0x018, %f6))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x018, %f6))
 	faligndata	%f2, %f4, %f18
-	EX_LD(LOAD(ldd, %o1 + 0x020, %f8))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x020, %f8))
 	faligndata	%f4, %f6, %f20
-	EX_LD(LOAD(ldd, %o1 + 0x028, %f10))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x028, %f10))
 	faligndata	%f6, %f8, %f22
 
-	EX_LD(LOAD(ldd, %o1 + 0x030, %f12))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x030, %f12))
 	faligndata	%f8, %f10, %f24
-	EX_LD(LOAD(ldd, %o1 + 0x038, %f14))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x038, %f14))
 	faligndata	%f10, %f12, %f26
-	EX_LD(LOAD(ldd, %o1 + 0x040, %f0))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x040, %f0))
 
 	subcc		GLOBAL_SPARE, 0x80, GLOBAL_SPARE
 	add		%o1, 0x40, %o1
@@ -184,26 +190,26 @@ FUNC_NAME:	/* %o0=dst, %o1=src, %o2=len */
 
 	.align		64
 1:
-	EX_LD(LOAD(ldd, %o1 + 0x008, %f2))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x008, %f2))
 	faligndata	%f12, %f14, %f28
-	EX_LD(LOAD(ldd, %o1 + 0x010, %f4))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x010, %f4))
 	faligndata	%f14, %f0, %f30
-	EX_ST(STORE_BLK(%f16, %o0))
-	EX_LD(LOAD(ldd, %o1 + 0x018, %f6))
+	EX_ST_FP(STORE_BLK(%f16, %o0))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x018, %f6))
 	faligndata	%f0, %f2, %f16
 	add		%o0, 0x40, %o0
 
-	EX_LD(LOAD(ldd, %o1 + 0x020, %f8))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x020, %f8))
 	faligndata	%f2, %f4, %f18
-	EX_LD(LOAD(ldd, %o1 + 0x028, %f10))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x028, %f10))
 	faligndata	%f4, %f6, %f20
-	EX_LD(LOAD(ldd, %o1 + 0x030, %f12))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x030, %f12))
 	subcc		%o3, 0x01, %o3
 	faligndata	%f6, %f8, %f22
-	EX_LD(LOAD(ldd, %o1 + 0x038, %f14))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x038, %f14))
 
 	faligndata	%f8, %f10, %f24
-	EX_LD(LOAD(ldd, %o1 + 0x040, %f0))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x040, %f0))
 	LOAD(prefetch, %o1 + 0x1c0, #one_read)
 	faligndata	%f10, %f12, %f26
 	bg,pt		%XCC, 1b
@@ -211,29 +217,29 @@ FUNC_NAME:	/* %o0=dst, %o1=src, %o2=len */
 
 	/* Finally we copy the last full 64-byte block. */
 2:
-	EX_LD(LOAD(ldd, %o1 + 0x008, %f2))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x008, %f2))
 	faligndata	%f12, %f14, %f28
-	EX_LD(LOAD(ldd, %o1 + 0x010, %f4))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x010, %f4))
 	faligndata	%f14, %f0, %f30
-	EX_ST(STORE_BLK(%f16, %o0))
-	EX_LD(LOAD(ldd, %o1 + 0x018, %f6))
+	EX_ST_FP(STORE_BLK(%f16, %o0))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x018, %f6))
 	faligndata	%f0, %f2, %f16
-	EX_LD(LOAD(ldd, %o1 + 0x020, %f8))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x020, %f8))
 	faligndata	%f2, %f4, %f18
-	EX_LD(LOAD(ldd, %o1 + 0x028, %f10))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x028, %f10))
 	faligndata	%f4, %f6, %f20
-	EX_LD(LOAD(ldd, %o1 + 0x030, %f12))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x030, %f12))
 	faligndata	%f6, %f8, %f22
-	EX_LD(LOAD(ldd, %o1 + 0x038, %f14))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x038, %f14))
 	faligndata	%f8, %f10, %f24
 	cmp		%g1, 0
 	be,pt		%XCC, 1f
 	 add		%o0, 0x40, %o0
-	EX_LD(LOAD(ldd, %o1 + 0x040, %f0))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x040, %f0))
 1:	faligndata	%f10, %f12, %f26
 	faligndata	%f12, %f14, %f28
 	faligndata	%f14, %f0, %f30
-	EX_ST(STORE_BLK(%f16, %o0))
+	EX_ST_FP(STORE_BLK(%f16, %o0))
 	add		%o0, 0x40, %o0
 	add		%o1, 0x40, %o1
 	membar		#Sync
@@ -253,20 +259,20 @@ FUNC_NAME:	/* %o0=dst, %o1=src, %o2=len */
 
 	sub		%o2, %g2, %o2
 	be,a,pt		%XCC, 1f
-	 EX_LD(LOAD(ldd, %o1 + 0x00, %f0))
+	 EX_LD_FP(LOAD(ldd, %o1 + 0x00, %f0))
 
-1:	EX_LD(LOAD(ldd, %o1 + 0x08, %f2))
+1:	EX_LD_FP(LOAD(ldd, %o1 + 0x08, %f2))
 	add		%o1, 0x8, %o1
 	subcc		%g2, 0x8, %g2
 	faligndata	%f0, %f2, %f8
-	EX_ST(STORE(std, %f8, %o0))
+	EX_ST_FP(STORE(std, %f8, %o0))
 	be,pn		%XCC, 2f
 	 add		%o0, 0x8, %o0
-	EX_LD(LOAD(ldd, %o1 + 0x08, %f0))
+	EX_LD_FP(LOAD(ldd, %o1 + 0x08, %f0))
 	add		%o1, 0x8, %o1
 	subcc		%g2, 0x8, %g2
 	faligndata	%f2, %f0, %f8
-	EX_ST(STORE(std, %f8, %o0))
+	EX_ST_FP(STORE(std, %f8, %o0))
 	bne,pn		%XCC, 1b
 	 add		%o0, 0x8, %o0
 
diff --git a/arch/sparc/lib/VISsave.S b/arch/sparc/lib/VISsave.S
index a063d8433..62c2647bd 100644
--- a/arch/sparc/lib/VISsave.S
+++ b/arch/sparc/lib/VISsave.S
@@ -6,24 +6,23 @@
  * Copyright (C) 1998 Jakub Jelinek (jj@ultra.linux.cz)
  */
 
+#include <linux/linkage.h>
+
 #include <asm/asi.h>
 #include <asm/page.h>
 #include <asm/ptrace.h>
 #include <asm/visasm.h>
 #include <asm/thread_info.h>
 
-	.text
-	.globl		VISenter, VISenterhalf
-
 	/* On entry: %o5=current FPRS value, %g7 is callers address */
 	/* May clobber %o5, %g1, %g2, %g3, %g7, %icc, %xcc */
 
 	/* Nothing special need be done here to handle pre-emption, this
 	 * FPU save/restore mechanism is already preemption safe.
 	 */
-
+	.text
 	.align		32
-VISenter:
+ENTRY(VISenter)
 	ldub		[%g6 + TI_FPDEPTH], %g1
 	brnz,a,pn	%g1, 1f
 	 cmp		%g1, 1
@@ -79,3 +78,4 @@ vis1:	ldub		[%g6 + TI_FPSAVED], %g3
 	.align		32
 80:	jmpl		%g7 + %g0, %g0
 	 nop
+ENDPROC(VISenter)
diff --git a/arch/sparc/mm/init_64.c b/arch/sparc/mm/init_64.c
index 4ac88b757..3025bd57f 100644
--- a/arch/sparc/mm/init_64.c
+++ b/arch/sparc/mm/init_64.c
@@ -93,6 +93,8 @@ static unsigned long cpu_pgsz_mask;
 static struct linux_prom64_registers pavail[MAX_BANKS];
 static int pavail_ents;
 
+u64 numa_latency[MAX_NUMNODES][MAX_NUMNODES];
+
 static int cmp_p64(const void *a, const void *b)
 {
 	const struct linux_prom64_registers *x = a, *y = b;
@@ -1157,6 +1159,48 @@ static struct mdesc_mlgroup * __init find_mlgroup(u64 node)
 	return NULL;
 }
 
+int __node_distance(int from, int to)
+{
+	if ((from >= MAX_NUMNODES) || (to >= MAX_NUMNODES)) {
+		pr_warn("Returning default NUMA distance value for %d->%d\n",
+			from, to);
+		return (from == to) ? LOCAL_DISTANCE : REMOTE_DISTANCE;
+	}
+	return numa_latency[from][to];
+}
+
+static int find_best_numa_node_for_mlgroup(struct mdesc_mlgroup *grp)
+{
+	int i;
+
+	for (i = 0; i < MAX_NUMNODES; i++) {
+		struct node_mem_mask *n = &node_masks[i];
+
+		if ((grp->mask == n->mask) && (grp->match == n->val))
+			break;
+	}
+	return i;
+}
+
+static void find_numa_latencies_for_group(struct mdesc_handle *md, u64 grp,
+					  int index)
+{
+	u64 arc;
+
+	mdesc_for_each_arc(arc, md, grp, MDESC_ARC_TYPE_FWD) {
+		int tnode;
+		u64 target = mdesc_arc_target(md, arc);
+		struct mdesc_mlgroup *m = find_mlgroup(target);
+
+		if (!m)
+			continue;
+		tnode = find_best_numa_node_for_mlgroup(m);
+		if (tnode == MAX_NUMNODES)
+			continue;
+		numa_latency[index][tnode] = m->latency;
+	}
+}
+
 static int __init numa_attach_mlgroup(struct mdesc_handle *md, u64 grp,
 				      int index)
 {
@@ -1220,9 +1264,16 @@ static int __init numa_parse_mdesc_group(struct mdesc_handle *md, u64 grp,
 static int __init numa_parse_mdesc(void)
 {
 	struct mdesc_handle *md = mdesc_grab();
-	int i, err, count;
+	int i, j, err, count;
 	u64 node;
 
+	/* Some sane defaults for numa latency values */
+	for (i = 0; i < MAX_NUMNODES; i++) {
+		for (j = 0; j < MAX_NUMNODES; j++)
+			numa_latency[i][j] = (i == j) ?
+				LOCAL_DISTANCE : REMOTE_DISTANCE;
+	}
+
 	node = mdesc_node_by_name(md, MDESC_NODE_NULL, "latency-groups");
 	if (node == MDESC_NODE_NULL) {
 		mdesc_release(md);
@@ -1245,6 +1296,23 @@ static int __init numa_parse_mdesc(void)
 		count++;
 	}
 
+	count = 0;
+	mdesc_for_each_node_by_name(md, node, "group") {
+		find_numa_latencies_for_group(md, node, count);
+		count++;
+	}
+
+	/* Normalize numa latency matrix according to ACPI SLIT spec. */
+	for (i = 0; i < MAX_NUMNODES; i++) {
+		u64 self_latency = numa_latency[i][i];
+
+		for (j = 0; j < MAX_NUMNODES; j++) {
+			numa_latency[i][j] =
+				(numa_latency[i][j] * LOCAL_DISTANCE) /
+				self_latency;
+		}
+	}
+
 	add_node_ranges();
 
 	for (i = 0; i < num_node_masks; i++) {
diff --git a/arch/sparc/net/bpf_jit_comp.c b/arch/sparc/net/bpf_jit_comp.c
index f8b9f71b9..3e6e05a7c 100644
--- a/arch/sparc/net/bpf_jit_comp.c
+++ b/arch/sparc/net/bpf_jit_comp.c
@@ -420,22 +420,9 @@ void bpf_jit_compile(struct bpf_prog *fp)
 		}
 		emit_reg_move(O7, r_saved_O7);
 
-		switch (filter[0].code) {
-		case BPF_RET | BPF_K:
-		case BPF_LD | BPF_W | BPF_LEN:
-		case BPF_LD | BPF_W | BPF_ABS:
-		case BPF_LD | BPF_H | BPF_ABS:
-		case BPF_LD | BPF_B | BPF_ABS:
-			/* The first instruction sets the A register (or is
-			 * a "RET 'constant'")
-			 */
-			break;
-		default:
-			/* Make sure we dont leak kernel information to the
-			 * user.
-			 */
+		/* Make sure we dont leak kernel information to the user. */
+		if (bpf_needs_clear_a(&filter[0]))
 			emit_clear(r_A); /* A = 0 */
-		}
 
 		for (i = 0; i < flen; i++) {
 			unsigned int K = filter[i].k;
@@ -812,7 +799,7 @@ cond_branch:			f_offset = addrs[i + filter[i].jf];
 	if (image) {
 		bpf_flush_icache(image, image + proglen);
 		fp->bpf_func = (void *)image;
-		fp->jited = true;
+		fp->jited = 1;
 	}
 out:
 	kfree(addrs);